Trust is needed to move from a non-cooperative to a cooperative situation, and it is clearly some kind of belief in "others" credibility. But, what exactly is trust? Here, I will limit myself to contrasting two different approaches. One, which has been presented by Russell Hardin, strives to keep the concept within the rationalistic instrumental paradigm (Hardin 1998). This means, A will trust B, if A believes that B's incentive structure is such that it is in B's interest to fulfill A's expectations in a particular exchange. According to Hardin, we do not trust people generally, but only in specific exchanges. For example, I do not trust my doctor to fix the brakes on my car.
As has been argued by Levi, the problem with this definition of trust is its narrowness (Levi 1999). As soon as B's incentive structure changes, (s)he will betray A. This puts a heavy burden on A, because he needs to know B's incentive structure very well, and also be aware of sudden changes in that incentive structure. The amount of information A needs in each an every moment, when A must decide whether of not to trust B, must be very high. While this definition is very precise, and has the advantage of simplicity, I doubt if it can capture trust between agents as it takes place in the real world. Do agents really make such complex calculations each an every time they decide whether or not to trust? Probably not, and in any case, the time and resources they would need to gather that type and amount of information about B, would make trust very rare (i.e., the transaction costs would prevent a lot of cooperative exchanges). Still, even in New York, in the middle of the night, people jump into a taxi, driven by a person they know nothing about (and that is one of the least dangerous thing that people do with complete strangers in this city). Furthermore, if according to Hardin, A will trust B as long as A assesses that it is in B's self-interest to act as to be trusted, then trust would be very infrequent because B would, following standard rational choice theory, pretend to be trustworthy but choose to "free-ride" on A's trust. This B personality would try to mimic trustworthiness until A decides to trust him with something really important or valuable, and then (s)he'd deceive A. It can also be argued that such an instrumental model of social ties leads to more fundamental problems. Granovetter has e.g., argued that "a perception of others that one's interest in them is mainly a matter of 'investment' will make this investment less likely to pay off; we are all on the lockout for those who only want to use us" (Granovetter 1988, p. 115). As Jon Elster have argued, there are willings that can not be willed (Elster 1989).
Definitions must not only be precise, elegant and simple (as is Hardin's), they must capture the essence of what we want to communicate when we use a specific term. As Robert Wuthnow concluded from his thorough empirical research about trust based on survey data: "For most people, trust is not simply a matter of making rational calculations about the possibility of benefiting by cooperating with someone else. Social scientist who reduce the study of trust to questions about rational choice, and who argue that it has nothing to do with moral discourse, miss that point" (Wuthnow 1998). The result from experimental research also seem to show that the calculative notion of trust does not conform to empirical findings (Rapoport 1987; Tyler and Degoey 1996). Recent overviews of the field states that experimental research simply refutes the behavioral assumptions of game theory and rational choice theory (Ledyard 1995; Sally 1995). Psychological research shows that trust is linked with norms against opportunism such as lying, cheating and stealing (Rotter 1980). At the end of the day, there must be limits to how unrealistic assumptions social scientists can work from and still pretend to have something important to say about real world problems.
As Ostrom has argued, we are in desperate need of a more realistic behavioral theory of human agency (Ostrom 1998). Using results from psychological research, Eckel and Wilson have stated that "it might be argued that the rational agents in standard game theoretic models are autistic: that is, they only require an actor to assume the other is seeking the same advantage as himself" (Eckel and Wilson 1999). They refer to psychological experiments with children which show that they can usually detect signals from other persons and make inferences about their intentions. However, children suffering from autism have difficulties doing this, and usually can not detect intentions other than their own which complicates their ability in social exchanges (cf. Baron-Cohen 1995). This argument may sound a bit harsh with respect to game theory, yet the argument is very compelling. As economists Ben-Ner and Putterman have recently argued:
being, by assumption, bereft of concern for friend and foe as well as for right or wrong, and caring only about his own well-being, homo economicus cannot, by construction, be at the center of a meaningful theory of how and when behavior is influenced by ethics, values, concern for others, and other preferences that depart from those of standard economic models. (Ben-Ner and Putterman 1998a).
Thus, there are two problems that need to be solved. One is how to incorporate norms into our models of human behavior. The other, is how to build our models from a somewhat more realistic assumption about the agents capability to handle and process information and act consistently upon this information. The first problem can be handled by thinking of agents as having "dual" utility functions in social dilemma situations. That is, they want to "do the right thing", that is, abstain from opportunistic behavior, but they do not want to be the "only ones" who are virtuous, because there is usually no point in being the only one who is virtuous. As a result of the information that others will only act according to their "myopic" self interest, they are likely to do the same (i.e., autistic action). But, if agents have information that "the others" have a normative orientation (or some other reason) that makes it likely that they will cooperate for the common good, the norm-based utility function will usually "kick in" (Levi 1991). Thus, what we want is agents that are motivated by self-interested when they act in market relations, but whom, when elected or appointed to a public position, for example as a judge or tax officials, obey the laws rather than ask for bribes (Ben-Ner and Putterman 1998b, p. 5)
In regard to the problem of using a more realistic approach about human behavior than the "autistic" standard model, I think that Peyton Young's approach on modeling agency in evolutionary game theory may be very promising. His conception of how we should understand human agency is that:
agents are not perfectly rational and fully informed about the world in which they live. They base their decisions on fragmentary information, they have incomplete models of the process they are engaged in, and they may not be especially forward looking. Still, they are not completely irrational: they adjust to their behavior based on what they think other agents are going to do, and these expectations are generated endogenously by information about what other agents have done in the past (Young 1998, p. 6)
There are several advantages with this view of agents compared to the standard model. Most important is the realistic view of what type of (limited) information agents have, the importance agents give to history (i.e., experience) when judging other agents, and that what they finally do, is based on what they think "other agents are going to do". A reasonable interpretation of the results from experimental studies and field studies about behavior in social dilemma situations, is that it is very likely agents do not only take into consideration the incentive structure of the "other agents". On the contrary, they are likely to take into account what is known about the moral standard, professional norms, and historical record of these "other agents". For example, do civil servants take bribes or not, do professors usually discriminate people of different race or gender or not, do judges the follow or violate legal rules, do union leaders honor agreements, or not, etc. In other words, what is the "logic of appropriatedness" of the other actors (March and Olsen 1989).
The question then, is where does this view of "the others" come from? In small groups it can of course come from personal knowledge and communication. Experimental research gives strong support for the importance of communication as a way of increasing cooperation in social dilemmas (Sally 1995). But, in Large N settings, the problem is very different. No agent can have knowledge about all the other agents or groups of agents. Who are the judges, the tax-collectors, the policemen, the politicians, the civil servants, the social welfare workers, the other tax payers, the welfare clients, the unions, etc? What sort of people are they? What is known about their trustworthiness and their moral standards? Can these people be trusted, and if so, with what can they be trusted? If agents "adjust to their behavior based on what they think other agents are going to do" (Young 1998, p. 6), these are the type of questions agents, who think about whether to trust or not, must try to answer. The stakes, if trust is misplaced, can be very high. As Braithwaite has stated "(t)he key to linking value systems and trust norms lies in the way in which 'the other' is construed" (Braithwaite 1998, p. 45).
Clearly, the questions raised by our Russian civil servant can not be answered within rational choice or game theory. These are very powerful analytical tools to help us state the problem of collective action and social dilemmas, but as argued above, they are less likely to help us understand why they are solved (Ostrom 1998). Miller and Hammond, for example, refer to the use of the so-called "city managers" as a successful way to get rid of corrupt party-machine politics in American cities during the interwar period. These where highly trained civil servants known for their high moral standards and for being disinterested, selfless servants of the public good. They had a reputation, that as a rule, they could not be bribed. But as Miller & Hammond state, "(t)o the extent that such a system works, it is clearly because city managers have been selected and/or trained not be economic actors" (Miller and Hammond 1994, p. 23). And, of course, there is then no collective action problem in the first place, because it is "solved" by blurring the assumption about human behavior on which the model is built. Their advice to our Russian friend mentioned above is, according to Miller and Hammond very simple: "to find out how such disinterested altruistic actors are created, and then reproduce them throughout the political system" (Miller and Hammond 1994, p. 24)). Well, what more can you say, than "good luck". The insight that such a non-cooperative equilibrium is known to be very robust, will not comfort him.
Where does social trust come from? The most important lessons so far are is that a) information is a problem and b) agents are likely to act based on what information they have about "the others." In certain economic setting, such as a spot market, information is free, general, accurate and immediately available. No agent can hide what he or she is willing to pay or to sell for from other agents. In politics, things are usually very different. Information is not free, almost never generally available and seldom accurate. Yet, agents have to act on what information they have (Mailath 1998). This information problem creates a demand for "information entrepreneurs". This is usually political leaders or intellectuals who engage in producing or reproducing ideas and system of ideas (i.e., ideology). What they do, to a large extent, is to produce ideas about trust. In a potential social dilemma situation, what are the actors' interests can not be taken for granted, because it depends on what ideas they think other actors have. In her work about the importance of ideas in major policy choices made by different European Social Democratic parties in the inter-war period, Berman argues that "actors with different ideas will make different decisions, even when placed in similar environment" (Berman 1998, p. 33). From a rational choice perspective, the argument is similar: "ideas matter because they affect how individuals interpret their world via the likelihood they accord different possibilities" (Bates, de Figueiredo Jr, and Weingast 1998).
If so, the solution to our problem must be found in who controls what information (or ideas) agents will use when deciding how to act (Berman 1998, ch. 2). The problems are, 1) where does ideas come from and, 2) what determines which ideas that will dominate. As Mailath puts it in his discussion about the realism in the approach known as evolutionary game theory: "(t)he consistency in Nash equilibrium seems to require that players know what the other players are doing. But where does this knowledge come from?" (Mailath 1998). One answer, of course, is that this is given by the "Culture", or the "Dominant Ideology", or "History". It is in the American political culture to "hate the government", while Scandinavians, for example, put enormous trust in their political system and gladly pay more than half of their income in taxes (Rothstein 1998). Russians have of course very good reasons to be distrustful of their government institutions, as do most citizens of Latin Europe and Latin America. According to Putnam (Putnam 1993), the reason Northern Italians trust each other can be traced back to political traditions established in the medieval city states, while southern Italians have much less social capital, because no such "horizontal" political culture ever got started3.
The problems with all these explanations are well-known. First, they reduce the agents to, more or less, cultural or structural "dopes" (Giddens 1984). Instead of being able to chose, they have no choice at all, and instead of having perfect information, they are "doped" by the culture in which, by no choice of their own, happen to live in. Second, cultural explanations have difficulties in getting at our main problem, namely how to explain change. If the culture is strong, then change in values is less likely. But if the culture is weak, change is possible but then other things than culture come into play (Laitin 1988).
This is the difficulty, in order to explain why social dilemmas can sometimes be solved, rationalistic theories must be completed with theories about how the agents come to embrace norms, ideas, or culture that make them refrain from self-defeating myopic instrumental "rational" behavior (Denzau and North 1994). But this must not lead us into the other extreme, i.e., viewing agents as determined by a cultural hegemony produced by anonymous historical or political forces (cf. Lichbach 1997). The whole notion of "social dilemmas" serves to remind us that groups do not always have the norms that would be most functional for theirs needs or interests. On the contrary, the rational choice approach's focus on the discrepancy between individual and collective rationality, implies that we should usually expect the opposite to be the case. Norms or culture should thus not be seen as something inherited and stable beyond strategic action. Instead, as Bates et. al have argued, "the struggle over subjective worldviews should itself be treated as a strategic process" (Bates, de Figueiredo Jr, and Weingast 1998).
This is were real-world politics kicks in, for the simple reason that it is very much a struggle over which "subjective worldviews" that shall dominate in a certain group (cf. Hardin 1995). Recent work by, for example, Berman and McNamara, about the importance of dominant ideas in major policy choices, point in the right direction, because which ideas that will dominate are fought about by strategically acting political agents. (Berman 1998; McNamara 1998). What political leaders do, is very much trying to communicate notions of who are "the others", and most especially, if these "others" can be trusted (Bates, de Figueiredo Jr, and Weingast 1998). "Others" can be ethnic groups or nations (the Serbs, the Croatians, the Tutsis), professional groups, "the bureaucrats in Washington", "the unions", "the employers", and so on. And within this strategic approach, when providing answers to the type of questions above, political leaders are at the same time providing an answer to the question of the identity of the group they represent: the "who are we" question (Ringmar 1996).
In the more empirical research, there are basically two arguments about how trust between citizens is produced. The first one comes from Robert Putnam's already classical study about variations between the Italian regions democratic efficiency. The main thrust of his argument is that "what makes democracy work" is trust, or social capital, and this is produced if citizens engage in horizontal voluntary organizations such as choral societies, PTAs and charities. In this Durkheimian notion of the social order, it is a vibrant civil society where citizens' engage in local grass root organizations, that they learn the noble art overcoming social dilemmas. At the aggregate level, Putnam is able to show very impressive correlations between the densitity of the world of voluntary organizations and democratic efficiency (Putnam 1993). There is also an argument that countries which historically have had strong popular grass-root organizations, such as the Scandinavian countries and the Netherlands, score high on the survey question used in the World Value studies to measure generalized trust (Inglehart 1997). At the micro-level, however, things look a little different. While there seem to be empirical support for the thesis that the more organizations people are members of, the more likely are they to trust other citizens, it is difficult to get a grip at how the causal relations works (Rothstein 1999). Is it agents who already trust other citizens who join many organizations, or is it the activity in the organizations that increases trust? Recent work by Dietlind Stolle seems to confirm the former thesis more than the latter. From her very interesting micro-level data, she concludes that "(p)eople who join associations are significantly more trusting than people who do not join", and "(i)t is not true that the longer and the more one associates, the greater one's generalized trust" (Stolle 1998, p. 521).
The other major argument about trust among citizens is that it can also be created "from above". Political and legal institutions that are perceived as fair, just and (reasonably) efficient, increase the likelihood that citizens will overcome social dilemmas (Levi 1998a; Rothstein 1998). If our friend, the Russian tax official, could make Russians believe that his tax bureaucrats would be honest and that they would have means to make sure that (almost) all other citizens paid their taxes, most Russians would pay there taxes. It should be added that although Putnam's analysis by most commentators has been connected to the "organic Durkheimian" understanding of trust, there are important parts of his book that take this more institutionalized causal relation into consideration. Be that as it may, Swedish survey data seem to give some support to this "statist" argument about how trust is created.
People were asked whether they had very high, high, middle, low or very low trust/confidence in different political and societal institutions such as the banks, Parliament, the unions, the police, the courts, etc. (see table below). The measure on this five-scale question of trust in institutions was then correlated with a ten-scale question asking people: "in general, do you think you can trust other people". The question we wanted to get at was whether there is any correlation between horizontal trust (i.e., trust in other people) and vertical trust (i.e., trust in political and societal institutions). Table 1 below shows the correlations between these two types of trust:
Table 1. Correlation Between Horizontal and Vertical Trust